Serious hypoglycemic reactions are one of the most common adverse drug events and are now recognized as a major public health issue. About 750,000 veterans and millions of Americans with diabetes are at high risk for serious hypoglycemia. From the mid-2000s to mid-2012, non-VA guidelines emphasized tight control of glycemia for most patients. Our recent publication reported that about half of Veterans Health Administration (VHA) patients at high risk of hypoglycemia were potentially over treated (A1c <7%). Our proposed overtreatment measure has been accepted as a concept measure for future development by Center for Medicare and Medicaid (CMS), following endorsement by the Department of Health and Human Services Office of the National Coordinator for Health Information Technology. In January 2015 VA will launch a Hypoglycemic Safety Initiative in the context of the Choosing Wisely Campaign. This study uniquely provides an administrative data base approach to tracking prevention and surveillance of emergency room visits and hospitalizations for hypoglycemia. Project Objectives: Aim 1. Assess the changes in rates and trends of potential overtreatment among various subgroups of high risk patients receiving insulin and/or sulfonylurea treatment during the period 2004-2016 in the context of changing guidelines. Aim 2. Evaluate rates, trends, and correlates of hypoglycemia resulting in emergency room visits or hospitalization in the years of 2004-2016 among high risk patients; Aim 3. Study rates and trends of mortality in the years of 2004-2016 among patients by insulin use; assess insulin, comorbidities, hypoglycemia, and A1c values as risk factors for mortality. Aim 4. Explore de-intensification of glycemic control among potentially over treated patients. Project Methods: This study includes both serial cross-sectional (Aims 1, 2, 3) and longitudinal (Aims 2, 3, 4) analyses of VHA Veteran patients based on their electronic health records from years 2002-2016. The dependent variables include overtreatment measures (<7%, <6.5%, and <6%), hypoglycemia resulting in emergency room visits or hospitalization, mortality, and medication de-intensification in glycemic control. Independent variables include patient characteristics and clinical factors (including but not limited to comorbidities, medications, body mass index, hospitalizations). Data sources include VHA and CMS data files. Data analysis will be conducted separately for veterans >=65 and <65 years old. All analyses will begin with inspection of frequency distributions, bivariate analyses o rates among subgroups defined by independent variables, and use of graphical methods for exploratory data analysis. Advanced statistical modeling will include marginal-Generalized estimating equations models with a logit link, Cox proportional hazards models, and the recursive partition classification tree method.